Improved Bandwidth and Noise Resilience in Thermal Impedance Spectroscopy by Mixing PRBS Signals

This paper presents a method of mixing pseudo-random binary sequences (PRBSs) to form a new signal that can be used to obtain the thermal impedance spectrum of power electronic systems. The proposed technique increases the useful frequency range of a PRBS by mixing two identical sequences at different frequencies. The new signal incorporates the frequency responses of both contributions. Mixing can be performed using a number of mathematical operators and analysis reveals that AND is the operator of choice since it has the lowest average input power for the same effectiveness. The bandwidth, frequency-domain representation, and noise resilience of PRBS signals are also reported. It is shown that the noise floor is significantly reduced under the mixed technique, which allows lower impedances to be measured under noisy measurement conditions. For a typical 8-bit PRBS, mixing reduces the noise floor by a factor of 10.5. Simulated and experimental validation are performed and results show the mixed scheme offers increased bandwidth, reduced computation and improved noise resilience compared to single PRBS techniques.